const char* device);

// The calls to TF_AddInput and TF_AddInputList must match (in number,
// order, and type) the op declaration.  For example, the "Concat" op
// has registration:
//   REGISTER_OP("Concat")
//       .Input("concat_dim: int32")
//       .Input("values: N * T")
//       .Output("output: T")
//       .Attr("N: int >= 2")
//       .Attr("T: type");

        interface.
        *   Add custom ops and kernels through
            [kernel and op registration C API](https://github.com/tensorflow/community/blob/master/rfcs/20190814-kernel-and-op-registration.md).
        *   Register custom graph optimization passes with

* Make logdump support kubemark and support gke with 'use_custom_instance_list' ([#51834](https://github.com/kubernetes/kubernetes/pull/51834), [@shyamjvs](https://github.com/shyamjvs))
* Fixes an issue with APIService auto-registration affecting rolling HA apiserver restarts that add or remove API groups being served. ([#51921](https://github.com/kubernetes/kubernetes/pull/51921), [@liggitt](https://github.com/liggitt))

- Added RuntimeClass restrictions and defaulting to PodSecurityPolicy. ([#73795](https://github.com/kubernetes/kubernetes/pull/73795), [@tallclair](https://github.com/tallclair))
- Kubelet plugin registration now has retry and exponential backoff logic for when registration of plugins (such as CSI or device plugin) fail. ([#73891](https://github.com/kubernetes/kubernetes/pull/73891), [@taragu](https://github.com/taragu))

([#64780](https://github.com/kubernetes/kubernetes/pull/64780), [@luxas](https://github.com/luxas))

* Add readyReplicas to replica sets ([#29481](https://github.com/kubernetes/kubernetes/pull/29481), [@kargakis](https://github.com/kargakis))

[Special Interest Groups]: https://github.com/kubernetes/community/blob/master/sig-list.md

### SIG API Machinery

[SIG API Machinery][] is responsible for all aspects of the API server: API registration and discovery, generic API CRUD semantics, admission control, encoding/decoding, conversion, defaulting, persistence layer (etcd), OpenAPI, third-party resources, garbage collection, and client libraries.

* The implicit registration of Prometheus metrics for workqueue has been removed, and a plug-able interface was added. If you were using workqueue in your own binaries and want these metrics, add the following to your imports in the main package: "k8s.io/pkg/util/workqueue/prometheus"....

- Kubelet: added retry of checking Unix domain sockets on Windows nodes for the plugin registration mechanism. ([#110075](https://github.com/kubernetes/kubernetes/pull/110075), [@luckerby](https://github.com/luckerby))

- Client-go: fixed the paged list calls with `ResourceVersionMatch` set would fail once paging is kicked in. ([#107311](https://github.com/kubernetes/kubernetes/pull/107311), [@fasaxc](https://github.com/fasaxc))
- Correct event registration for multiple scheduler plugins; this fixes a potential significant delay in re-queueing unschedulable pods. ([#109442](https://github.com/kubernetes/kubernetes/pull/109442), [@ahg-g](https://github.com/ahg-g))